A Digital Speech and Language Sample Application for Clinical Diagnostics and Monitoring in Swiss German (DiSpLa)

We are developing the first clinically usable platform for automated speech and language analysis in Swiss German. By combining advanced speech recognition with language and voice biomarkers, the system will support faster, more accurate diagnosis and monitoring of disorders such as developmental language disorders and dementia.

Key data

Description

Many clinical populations, such as children with developmental language disorders (DLDs) and older adults with dementia, show characteristic speech and language impairments. Diagnosing and monitoring these conditions typically relies on Language Sample Analysis (LSA), a method that is highly time- and cost-intensive due to the need for manual transcription and annotation.

In Swiss German, a dialect with limited digital resources, the lack of suitable automatic speech recognition (ASR) Tools further restricts the feasibility of automated LSA.

We aim to develop the first application that integrates advanced Swiss German ASR with Extended speech and language analysis. The tool will preserve speaker errors, handle child and impaired speech, and incorporate clinically relevant speech markers beyond traditional LSA. The project builds on Prior work in error-preserving STT models, Swiss German ASR, and clinical research on speech and voice markers in dementia and DLDs.

Work will be organized in four packages:

Development of a robust Swiss German ASR system
Implementation of a secure application,
Integration of speech and voice biomarkers, and
Transfer into clinical practice via testing, workshops, and public dissemination.

If successful, the project will reduce the time and cost of LSA by an order of magnitude, increase diagnostic accuracy, and provide
the first clinically usable speech and language analysis platform for Swiss German.